A self-hosted chat interface for AI models, shared by the whole team: accounts and roles, chat history, document and knowledge-base search (RAG), web search and tools. Connects to any OpenAI-compatible API (OpenAI, Anthropic via a gateway, OpenRouter, Groq…) or an Ollama server you run elsewhere; no model is bundled. Runs as a single container: SQLite, uploaded files and the embedding model cache live in one persistent volume. The first account created becomes the administrator. Above 50 users the licence requires the Open WebUI branding to stay visible.
The easiest way is through the Console: connect a server, then install the app from the catalog.
Sign in to the Keenrig Console
Add a server and run the install command the Console gives you on your VPS
Go to Catalog, pick this app and press Install
Open WebUI is a ChatGPT-style chat interface your whole team shares on your own machine: each person gets an account and their own chat history, and the administrator decides who can use which model. It is not a model itself — it connects to an API you already have (OpenAI, OpenRouter, Groq or any OpenAI-compatible service) or to an Ollama server, so you pay for usage instead of per seat.
What it does not do. This entry ships no AI model: without an API key or an Ollama server there is nothing to chat with. Running models on this VPS is not practical without a GPU. It needs at least 2 GB of memory because the embedding model for document search runs inside the container, and the first start downloads that model, so it takes a few minutes. Create the admin account right after installing: the first person to sign up becomes the administrator. Above 50 users, the Open WebUI licence requires its branding to stay visible.
Maintain Open WebUI? Paste this into your README so readers can install it with Keenrig:
[](https://keenrig.com/en/apps/open-webui)Compared with other tools: Compared with Cloudron · Compared with Coolify